Tsan-Hung Fu’s journal paper has been officially published in the Journal of Manufacturing Processes. This study proposes a novel vision-based, sensor-free framework for real-time error detection and correction in material extrusion (MEX) additive manufacturing. Using neural networks trained on camera images, the system predicts process parameters with high precision, eliminating the need for extra sensors. It integrates dynamic monitoring and corrective pause mechanisms, reducing defects to under 5 mm and achieving rapid response times of under 30 seconds. Experiments on various geometries confirmed consistent error correction, preservation of mechanical properties, and strong adaptability to unforeseen defects. This approach offers a scalable solution for improving process stability, efficiency, and quality in industrial additive manufacturing.
傅粲閎同學的期刊論文已正式發表於《Journal of Manufacturing Processes》。傅同學的研究題目為”Real-time process monitoring and error correction in material extrusion-based additive manufacturing via multi-output machine learning”. 本研究提出一種新型的視覺化、無感測器框架，用於材料擠製（MEX）積層製造的即時錯誤偵測與修正。系統透過以相機影像訓練的神經網路高精度預測製程參數，無需額外感測器。它結合動態監測與暫停修正機制，能將缺陷縮小至 5 公釐以下，並在 30 秒內快速回應。針對多種幾何形狀的實驗驗證顯示，該系統可穩定修正錯誤、保持機械性能，並具備處理未知缺陷的強大適應性。此方法為提升工業積層製造的製程穩定性、效率與品質，提供了一個可擴展的解決方案。